The invention provides an X-
ray-based
scoliosis 3D intelligent automatic diagnosis
system, which is characterized in that a spine area bounding frame coordinate, a
hip bone bounding frame coordinate and a
vertebral pedicle bounding frame coordinate are extracted from spine coronal image data, and the first 17 results are extracted to be matched with the number and
spatial distribution of typical vertebral bodies of a
human body; the method comprises the following steps: obtaining a spine region sub-graph, a
hip bone region sub-graph and a
vertebral pedicle region sub-graph, converting the sub-graphs into tensors to obtain a bounding box, four
angular point coordinates and corresponding confidence of each
vertebral body, storing the bounding box, the four
angular point coordinates and the corresponding confidence as a structured
record, performing Cobb angle calculation according to the structured
record, and converting the structured
record into an anatomical segment identifier;
cutting is conducted according to the spine area, the
hip bone area and the
vertebral pedicle area subgraph, segmented vertebral pedicles and hip bones are obtained, the major curvature type, the
pelvis height and the rotation direction are judged according to anatomical segment identifiers, and the
patient age is determined according to the midpoints of the upper edges and the lower edges of the vertebral bodies, the arrangement angle of the adjacent vertebral bodies, spine major curvature core parameters and the age of the patient. And outputting a
scoliosis 3D intelligent automatic diagnosis result.